musia-song-localization

Pass

Audited by Gen Agent Trust Hub on Oct 19, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes several CLI tools including conda, python, bash, and npm. These are used to run audio analysis pipelines, generate localization packages, and manage backend environments.
  • [EXTERNAL_DOWNLOADS]: The workflow includes instructions to run download_quality_backends.sh to fetch external model weights and dependencies required for singing synthesis.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes lyric text and audio transcriptions provided as external files.
  • Ingestion points: Reads target lyrics from TARGET_LINES.txt and source segments from lyrics.json in the create_localization_pack.py script.
  • Boundary markers: Includes non-negotiable rules for manual listening and a "missing-planned-phrase audit" to ensure accuracy against the audible vocal.
  • Capability inventory: Performs file system writes (creating JSON/text artifacts) and invokes shell-based audio processing.
  • Sanitization: Lacks explicit programmatic sanitization for the content of processed lyrics.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 19, 2026, 06:48 PM